DOCSLIB.ORG

CS 268: Structured P2P Networks: Pastry and Tapestry

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
CS 268: Structured P2P Networks: Pastry and Tapestry Domain CS 268: Structured peer-to-peer overlay networks - Sometimes called DHTs, DOLRs, CASTs, … Structured P2P Networks: - Examples: CAN, Chord, Pastry, Tapestry, … Pastry and Tapestry Contrasted with unstructured P2P networks - Gnutella, Freenet, etc. Today talking about Pastry and Tapestry Sean Rhea April 29, 2003 2 Service Model Service Model (con’t.) Let Ι be a set of identifiers Owner mapping exposed in variety of ways - Such as all 160-bit unsigned integers - In Chord, have function Let Ν be a set of nodes in a P2P system n = find_successor (i) - Some subset of all possible (IP, port) tuples - In Pastry and Tapestry, have function Structured P2P overlays implement a mapping route_to_root (i,m) ¡ In general, can be iterative or recursive owner Ν: Ι Ν - Iterative = directed by querying node - Given any identifier, deterministically map to a node - Recursive = forwarded through network Properties ¡ May also expose owner –1: Ν P(Ι) - Should take O(log |Ν|) time and state per node - Which identifiers given node is responsible for - Should be roughly load balanced 3 4 1 Other Service Models Lecture Overview Other models can be implemented on owner Introduction Example: Distributed hash table (DHT) PRR Trees - Overview void put (key, data) { - Locality Properties Pastry n = owner (key) - Routing in Pastry n.hash_table.insert (key, data) - Joining a Pastry network } - Leaving a Pastry network Tapestry data get (key) { - Routing in Tapestry n = owner (key) - Object location in Tapestry return n.hash_table.lookup (key) Multicast in PRR Trees } Conclusions 5 6 PRR Trees PRR Trees: The Basic Idea Work by Plaxton, Rajaraman, Richa (SPAA ’97) Basic idea: add injective function ¡ - Interesting in a distributed publication system node_id: Ν Ι - Similar to Napster in interface - Gives each node a name in the identifier space - Only for static networks (set Ν does not change) owner (i) = node whose node_id is “closest” to i Ι - No existing implementation (AFAIK) - Definition of closest varies, but always over Pastry and Tapestry both based on PRR trees To find owner (i) from node with identifier j 1. Let p = longest matching prefix between i and j - Extend to support dynamic node membership 2. Find node k with longest matching prefix of |p|+1 digits - Several implementations 3. If no such node, j is the owner (root) 4. Otherwise, forward query to node k Step 2 is the tricky part 7 8 2 PRR Trees: The Routing Table PRR Trees: Routing Ν Each node n has O(b log b | |) neighbors To find owner (47E2) - Each Lx neighbor shares x digits with n - Query starts at node 3AF2 - Set of neighbors forms a routing table - Resolve first digit by routing to 4633 - Resolve second digit by routing to 47DA, etc. 3A01 9CD0 3AFC L2 L0 L3 2974 45B3 47C1 3AF2 3AF2 4633 47DA 47EC L1 L3 L0 3C57 5A8F 4889 47F7 443E 3AF1 9 10 PRR Trees: Routing (con’t.) PRR Trees: Handling Inexact Matches Problem: what if no exact match? Want owner function to be deterministic - Consider the following network - Must have a way to resolve inexact matches - Who is the owner of identifier 3701? Solved different ways by each system Network is well formed 1000 - I have no idea what PRR did - Every routing table spot that can be - Pastry chooses numerically closest node filled is filled • Can break ties high or low 2000 3800 - Can route to all node identifiers - Tapestry performs “surrogate routing” Owner of 3701 not well defined • Chooses next highest match on per digit basis - Starting from 1000, it’s node 3800 3600 More on this later - Starting from 2000, it’s node 3600 Violation of service model 11 12 3 Locality in PRR Trees Locality in PRR Trees: Experiments Consider a node with id=1000 in a PRR network - At lowest level of routing table, node 1000 needs neighbors with prefixes 2-, 3-, 4-, etc. - In a large network, there may be several of each Idea: chose the “best” neighbor for each prefix - Best can mean lowest latency, highest bandwidth, etc. Can show that this choice gives good routes - For certain networks, routing path from query source to owner no more than a constant worse than routing path in underlying network - I’m not going to prove this today, see PRR97 for details 13 14 Lecture Overview Pastry Introduction Introduction A PRR tree combined with a Chord-like ring PRR Trees - Each node has PRR-style neighbors - Overview - And each node knows its predecessor and successor - Locality Properties Pastry • Called its leaf set - Routing in Pastry To find owner (i), node n does the following: - Joining a Pastry network - If i is n’s leaf set, choose numerically closest node - Leaving a Pastry network - Else, if appropriate PRR-style neighbor, choose that Tapestry - Routing in Tapestry - Finally, choose numerically closest from leaf set - Object location in Tapestry A lot like Chord Multicast in PRR Trees - Only leaf set necessary for correctness Conclusions - PRR-neighbors like finger table, only for performance 15 16 4 Pastry Routing Example Notes on Pastry Routing PRR neighbors in black Leaf set is great for correctness Leaf set neighbors in blue - Need not get PRR neighbors correct, only leaf set Owner of 3701 is now well-defined - If you believe the Chord work, this isn’t too hard to do From 1000 1000 Leaf set also gives implementation of owner -1(n) - Resolve first digit routing to 3800 - All identifiers half-way between n and its predecessor to - At 3800, see that we’re done half-way between n and its successor - (Numerically closer than 3600) 2000 3800 Can store k predecessors and successors From 2000 - Resolve first digit routing to 3600 - Gives further robustness as in Chord - At 3600, 3701 is in leaf set 3600 • In range 2000-3800 - Route to 3800 b/c numerically closer 17 18 Joining a Pastry Network Pastry Join Example Must know of a “gateway” node, g Node 3701 wants to join Join path Ι - Has 1000 as gateway Pick new node’s identifier, n, U.A.R. from ¡ Join path is 1000 3800 1000 Ask g to find the m = owner (n) - 3800 is the owner - And ask that it record the path that it takes to do so 3701 ties itself into leaf set 2000 3800 Ask m for its leaf set 3701 builds routing table Contact m’s leaf set and announce n’s presence - L0 neighbors from 1000 • 1000, 2000, and 3800 - These nodes add n to their leaf sets and vice versa 3600 - L1 neighbors from 3800 Build routing table • 3600 - Get level i of routing table from node i in the join path Existing nodes on join path 3701 - Use those nodes to make level i of our routing table consider 3701 as a neighbor 19 20 5 Pastry Join Notes Pastry Join Optimization Join is not “perfect” Best if gateway node is “close” to joining node - A node whose routing table needs new node should learn about it - Gateway joined earlier, should have close neighbors • Necessary to prove O(log |Ν|) routing hops - Recursively, gateway’s neighbors’ neighbors are close - Also not guaranteed to find close neighbors - Join path intuitively provides good initial routing table Can use routing table maintenance to fix both - Less need to fix up with routing table maintenance - Periodically ask neighbors for their neighbors Pastry’s optimized join algorithm - Use to fix routing table holes; replace existing distant neighbors - Before joining, find a good gateway, then join normally Philosophically very similar to Chord To find a good gateway, refine set of candidates - Start with minimum needed for correctness (leaf set) - Start with original gateway’s leaf set - Patch up performance later (routing table) - Keep only a closest few, then add their neighbors - Repeat (more or less--see paper for details) 21 22 Leaving a Pastry Network Dealing with Broken Leaf Sets Do not distinguish between leaving and crashing What if no nodes left in leaf set due to failures? - A good design decision, IMHO Can use routing table to recover (MCR93) Remaining nodes notice leaving node n down - Choose closest nodes in routing table to own identifier - Stops responding to keep-alive pings - Ask them for their leaf sets Fix leaf sets immediately - Choose closest of those, recurse - Easy if 2+ predecessors and successors known Allows use of smaller leaf sets Fix routing table lazily - Wait until needed for a query - Or until routing table maintenance - Arbitrary decision, IMHO 23 24 6 Lecture Overview Tapestry Routing Introduction Only different from Pastry when no exact match PRR Trees - Instead of using next numerically closer node, use node - Overview with next higher digit at each hop - Locality Properties Pastry Example: - Routing in Pastry - Given 3 node network: nodes 0700, 0F00, and FFFF - Joining a Pastry network - Who owns identifier 0000? - Leaving a Pastry network In Pastry, FFFF does (numerically closest) Tapestry - Routing in Tapestry In Tapestry, 0700 does - Object location in Tapestry - From FFFF to 0700 or 0F00 (doesn’t matter) Multicast in PRR Trees - From 0F00 to 0700 (7 is next highest digit after 0) Conclusions - From 0700 to itself (no node with digit between 0 and 7) 25 26 Notes on Tapestry Routing Object Location in Tapestry Mostly same locality properties as PRR and Pastry Pastry was originally just a DHT But compared to Pastry, very fragile - Support for multicast added later (RKC+01) Consider previous example: 0700, 0F00, FFFF PRR and Tapestry are DOLRs - What if 0F00 doesn’t know about 0700? - Distributed Object Location and Routing - 0F00 will think it is the owner of 0000 Service model - 0700 will still think it is the owner - publish (name) - Mapping won’t be deterministic throughout network - route_to_object (name, message) Tapestry join algorithm guarantees won’t happen - All routing table holes than can be filled will be Like Napster, Gnutella, and DNS - Provably correct, but tricky to implement - Service does not store data, only pointers to it - Leaf set links are bidirectional, easier to keep consistent - Manages a mapping of names to hosts 27 28 7 A Simple DOLR Implementation Problems with Simple DOLR Impl.
Load more

Recommended publications
  • A Content-Addressable Network for Similarity Search in Metric Spaces Fabrizio Falchi
    University of Pisa Masaryk University Faculty of Informatics Department of Department of Information Engineering Information Technologies Doctoral Course in Doctoral Course in Information Engineering Informatics Ph.D. Thesis A Content-Addressable Network for Similarity Search in Metric Spaces Fabrizio Falchi Supervisor Supervisor Supervisor Lanfranco Lopriore Fausto Rabitti Pavel Zezula 2007 Abstract Because of the ongoing digital data explosion, more advanced search paradigms than the traditional exact match are needed for content- based retrieval in huge and ever growing collections of data pro- duced in application areas such as multimedia, molecular biology, marketing, computer-aided design and purchasing assistance. As the variety of data types is fast going towards creating a database utilized by people, the computer systems must be able to model hu- man fundamental reasoning paradigms, which are naturally based on similarity. The ability to perceive similarities is crucial for recog- nition, classification, and learning, and it plays an important role in scientific discovery and creativity. Recently, the mathematical notion of metric space has become a useful abstraction of similarity and many similarity search indexes have been developed. In this thesis, we accept the metric space similarity paradigm and concentrate on the scalability issues. By exploiting computer networks and applying the Peer-to-Peer communication paradigms, we build a structured network of computers able to process similar- ity queries in parallel. Since no centralized entities are used, such architectures are fully scalable. Specifically, we propose a Peer- to-Peer system for similarity search in metric spaces called Met- ric Content-Addressable Network (MCAN) which is an extension of the well known Content-Addressable Network (CAN) used for hash lookup.
    [Show full text]
  • Tutorial on P2P Systems
    P2P Systems Keith W. Ross Dan Rubenstein Polytechnic University Columbia University http://cis.poly.edu/~ross http://www.cs.columbia.edu/~danr [email protected] [email protected] Thanks to: B. Bhattacharjee, A. Rowston, Don Towsley © Keith Ross and Dan Rubenstein 1 Defintion of P2P 1) Significant autonomy from central servers 2) Exploits resources at the edges of the Internet P storage and content P CPU cycles P human presence 3) Resources at edge have intermittent connectivity, being added & removed 2 It’s a broad definition: U P2P file sharing U DHTs & their apps P Napster, Gnutella, P Chord, CAN, Pastry, KaZaA, eDonkey, etc Tapestry U P2P communication U P Instant messaging P2P apps built over P Voice-over-IP: Skype emerging overlays P PlanetLab U P2P computation P seti@home Wireless ad-hoc networking not covered here 3 Tutorial Outline (1) U 1. Overview: overlay networks, P2P applications, copyright issues, worldwide computer vision U 2. Unstructured P2P file sharing: Napster, Gnutella, KaZaA, search theory, flashfloods U 3. Structured DHT systems: Chord, CAN, Pastry, Tapestry, etc. 4 Tutorial Outline (cont.) U 4. Applications of DHTs: persistent file storage, mobility management, etc. U 5. Security issues: vulnerabilities, solutions, anonymity U 6. Graphical structure: random graphs, fault tolerance U 7. Experimental observations: measurement studies U 8. Wrap up 5 1. Overview of P2P U overlay networks U P2P applications U worldwide computer vision 6 Overlay networks overlay edge 7 Overlay graph Virtual edge U TCP connection U or
    [Show full text]
  • Content Addressable Network (CAN) Is an Example of a DHT Based on a D-Dimensional Torus
    Overlay and P2P Networks Structured Networks and DHTs Prof. Sasu Tarkoma 6.2.2014 Contents • Today • Semantic free indexing • Consistent Hashing • Distributed Hash Tables (DHTs) • Thursday (Dr. Samu Varjonen) • DHTs continued • Discussion on geometries Rings Rings are a popular geometry for DHTs due to their simplicity. In a ring geometry, nodes are placed on a one- dimensional cyclic identifier space. The distance from an identifier A to B is defined as the clockwise numeric distance from A to B on the circle Rings are related with tori and hypercubes, and the 1- dimensional torus is a ring. Moreover, a k-ary 1-cube is a k-node ring The Chord DHT is a classic example of an overlay based on this geometry. Each node has a predecessor and a successor on the ring, and an additional routing table for pointers to increasingly far away nodes on the ring Chord • Chord is an overlay algorithm from MIT – Stoica et. al., SIGCOMM 2001 • Chord is a lookup structure (a directory) – Resembles binary search • Uses consistent hashing to map keys to nodes – Keys are hashed to m-bit identifiers – Nodes have m-bit identifiers • IP-address is hashed – SHA-1 is used as the baseline algorithm • Support for rapid joins and leaves – Churn – Maintains routing tables Chord routing I Identifiers are ordered on an identifier circle modulo 2m The Chord ring with m-bit identifiers A node has a well determined place within the ring A node has a predecessor and a successor A node stores the keys between its predecessor and itself The (key, value) is stored on the successor
    [Show full text]
  • Overlook: Scalable Name Service on an Overlay Network
    Overlook: Scalable Name Service on an Overlay Network Marvin Theimer and Michael B. Jones Microsoft Research, Microsoft Corporation One Microsoft Way Redmond, WA 98052, USA {theimer, mbj}@microsoft.com http://research.microsoft.com/~theimer/, http://research.microsoft.com/~mbj/ Keywords: Name Service, Scalability, Overlay Network, Adaptive System, Peer-to-Peer, Wide-Area Distributed System Abstract 1.1 Use of Overlay Networks This paper indicates that a scalable fault-tolerant name ser- Existing scalable name services, such as DNS, tend to rely vice can be provided utilizing an overlay network and that on fairly static sets of data replicas to handle queries that can- such a name service can scale along a number of dimensions: not be serviced out of caches on or near a client. Unfortu- it can be sized to support a large number of clients, it can al- nately, static designs don’t handle flash crowd workloads very low large numbers of concurrent lookups on the same name or well. We want a design that enables dynamic addition and sets of names, and it can provide name lookup latencies meas- deletion of data replicas in response to changing request loads. ured in seconds. Furthermore, it can enable updates to be Furthermore, in order to scale, we need a means by which the made pervasively visible in times typically measured in sec- changing set of data replicas can be discovered without requir- onds for update rates of up to hundreds per second. We ex- ing global changes in system routing knowledge. plain how many of these scaling properties for the name ser- Peer-to-peer overlay routing networks such as Tapestry vice are obtained by reusing some of the same mechanisms [24], Chord [22], Pastry [18], and CAN [15] provide an inter- that allowed the underlying overlay network to scale.
    [Show full text]
  • D3.5 Design and Implementation of P2P Infrastructure
    INNOVATION ACTION H2020 GRANT AGREEMENT NUMBER: 825171 WP3 – Vision Materialisation and Technical Infrastructure D3.5 – Design and implementation of P2P infrastructure Document Info Contractual Delivery Date: 31/12/2019 Actual Delivery Date: 31/12/2019 Responsible Beneficiary: INOV Contributing Beneficiaries: UoG Dissemination Level: Public Version: 1.0 Type: Final This project has received funding from the European Union’s H2020 research and innovation programme under the grant agreement No 825171 ` DOCUMENT INFORMATION Document ID: D3.5: Design and implementation of P2P infrastructure Version Date: 31/12/2019 Total Number of Pages: 36 Abstract: This deliverable describes the EUNOMIA P2P infrastructure, its design, the EUNOMIA P2P APIs and a first implementation of the P2P infrastructure and the corresponding APIs made available for the remaining EUNOMIA modules to start integration (for the 1st phase). Keywords: P2P design, P2P infrastructure, IPFS AUTHORS Full Name Beneficiary / Organisation Role INOV INESC INOVAÇÃO INOV Overall Editor University of Greenwich UoG Contributor REVIEWERS Full Name Beneficiary / Organisation Date University of Nicosia UNIC 23/12/2019 VERSION HISTORY Version Date Comments 0.1 13/12/2019 First internal draft 0.6 22/12/2019 Complete draft for review 0.8 29/12/2019 Final draft following review 1.0 31/12/2019 Final version to be released to the EC Type of deliverable PUBLIC Page | ii H2020 Grant Agreement Number: 825171 Document ID: WP3 / D3.5 EXECUTIVE SUMMARY This deliverable describes the P2P infrastructure that has been implemented during the first phase of the EUNOMIA to provide decentralized support for storage, communication and security functions. It starts with a review of the main existing P2P technologies, where each one is analysed and a selection of candidates are selected to be used in the project.
    [Show full text]
  • Effective Manycast Messaging for Kademlia Network
    Effective Manycast Messaging for Kademlia Network Lubos Matl Tomas Cerny Michael J. Donahoo Czech Technical University Czech Technical University Baylor University Charles square 13, 121 35 Charles square 13, 121 35 Waco, TX, US Prague 2, CZ Prague 2, CZ [email protected] [email protected] [email protected] ABSTRACT General Terms Peer-to-peer (P2P) communication plays an ever-expanding Measurement, Reliability, Performance, Experimentation role in critical applications with rapidly growing user bases. In addition to well-known P2P systems for data sharing Keywords (e.g., BitTorrent), P2P provides the core mechanisms in VoIP (e.g., Skype), distributed currency (e.g., BitCoin), etc. Distributed systems, performance, communication, P2P, There are many communication commonalities in P2P ap- DHT, key-based search plications; consequently, we can factor these communication primitives into overlay services. Such services greatly sim- 1. INTRODUCTION plify P2P application development and even allow P2P in- P2P networking is a widely-known architectural pattern frastructures to host multiple applications, instead of each for distributed systems. P2P has been applied to a variety of having its own network. Note well that such services must be applications, including file sharing (BitTorrent [10]), video both self-scaling and robust to meet the needs of large, ad- or audio streaming (SplitStream [5], Coolstream [22]), par- hoc user networks. One such service is Distributed Hash Ta- allel computation, online payment systems (Bitcoin [16]), ble (DHT) providing a dictionary-like location service, useful or voice-over-IP services (Skype). All these applications in many types of P2P applications. Building on the DHT demonstrate that the architecture is important and useful primitives for search and store, we can add even more power- for multiple domains.
    [Show full text]
  • One Ring to Rule Them All: Service Discovery and Binding in Structured Peer-To-Peer Overlay Networks
    One Ring to Rule them All: Service Discovery and Binding in Structured Peer-to-Peer Overlay Networks Miguel Castro Peter Druschel Anne-Marie Kermarrec Antony Rowstron Microsoft Research, Rice University, Microsoft Research, Microsoft Research, 7 J J Thomson Close, 6100 Main Street, 7 J J Thomson Close, 7 J J Thomson Close, Cambridge, MS-132, Houston, Cambridge, Cambridge, CB3 0FB, UK. TX 77005, USA. CB3 0FB, UK. CB3 0FB, UK. [email protected] [email protected] [email protected] [email protected] One Ring to rule them all. One Ring to find them. One Ring to bring In these systems a live node in the overlay to each key them all. And in the darkness bind them. J.R.R. Tolkien and provide primitives to send a message to a key. Mes- Abstract sages are routed to the live node that is currently responsi- ble for the destination key. Keys are chosen from a large space and each node is assigned an identifier (nodeId) cho- Self-organizing, structured peer-to-peer (p2p) overlay sen from the same space. Each node maintains a routing networks like CAN, Chord, Pastry and Tapestry offer a table with nodeIds and IP addresses of other nodes. The novel platform for a variety of scalable and decentralized protocols use these routing tables to assign keys to live distributed applications. These systems provide efficient nodes. For instance, in Pastry, a key is assigned to the live and fault-tolerant routing, object location, and load bal- node with nodeId numerically closest to the key. ancing within a self-organizing overlay network.
    [Show full text]
  • Distributed Systems – Lectures
    377 P2P Systems 11.1 Introduction & overview 11 Peer-to-Peer Systems1 11.1 Introduction & overview Client-server Model • Clients and servers each with distinct roles • The server and the network become the bottlenecks and points of failure Peer-to-peer Model “Peer-to-Peer (P2P) is a way of structuring distributed applications such that the individual nodes have symmetric roles. Rather than being divided into clients and servers each with quite distinct roles, in P2P applications a node may act as both a client and a server.” (Excerpt from the Charter of Peer-to-Peer Research Group, IETF/IRTF, June 24, 2003 http://www.irtf.org/charters/p2prg.html) 1Textbook Chapter 10 378 P2P Systems 11.1 Introduction & overview • Peers play similar roles • No distinction of responsibilities Key characteristics of P2P Systems: • Ensures that each user contributes resources to the system • All the nodes have the same functional capabilities and responsibilities • Their correct operation does not depend on the existence of any centrally- administered systems • They can be designed to offer a limited degree of anonymity to the providers and users of resources • A key issue: placement of data across many hosts – efficiency 379 P2P Systems 11.1 Introduction & overview – load balancing – availability Generations • Early services – DNS, Netnews/Usenet – Xerox Grapevine name/mail service – Lamport’s part-time parliament algorithm for distributed consesnus – Bayou replicated storage system – classless inter-domain IP routing algorithm • 1st generation – centralized
    [Show full text]
  • Tapestry: a Resilient Global-Scale Overlay for Service Deployment Ben Y
    IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 22, NO. 1, JANUARY 2004 41 Tapestry: A Resilient Global-Scale Overlay for Service Deployment Ben Y. Zhao, Ling Huang, Jeremy Stribling, Sean C. Rhea, Anthony D. Joseph, Member, IEEE, and John D. Kubiatowicz, Member, IEEE Abstract—We present Tapestry, a peer-to-peer overlay routing delivery to mobile or replicated endpoints in the presence of infrastructure offering efficient, scalable, location-independent instability in the underlying infrastructure. As a result, a DOLR routing of messages directly to nearby copies of an object or network provides a simple platform on which to implement service using only localized resources. Tapestry supports a generic decentralized object location and routing applications program- distributed applications—developers can ignore the dynamics ming interface using a self-repairing, soft-state-based routing of the network except as an optimization. Already, Tapestry has layer. This paper presents the Tapestry architecture, algorithms, enabled the deployment of global-scale storage applications and implementation. It explores the behavior of a Tapestry such as OceanStore [4] and multicast distribution systems such deployment on PlanetLab, a global testbed of approximately 100 as Bayeux [5]; we return to this in Section VI. machines. Experimental results show that Tapestry exhibits stable Tapestry is a peer-to-peer (P2P) overlay network that pro- behavior and performance as an overlay, despite the instability of the underlying network layers. Several widely distributed vides high-performance, scalable, and location-independent applications have been implemented on Tapestry, illustrating its routing of messages to close-by endpoints, using only localized utility as a deployment infrastructure.
    [Show full text]
  • 0 to 10K in 20 Seconds: Bootstrapping Large-Scale DHT Networks
    0 to 10k in 20 seconds: Bootstrapping Large-scale DHT networks Jae Woo Lee∗, Henning Schulzrinne∗, Wolfgang Kellerer† and Zoran Despotovic† ∗ Department of Computer Science, Columbia University, New York, USA {jae,hgs}@cs.columbia.edu † DoCoMo Communications Laboratories Europe, Munich, Germany {kellerer,despotovic}@docomolab-euro.com Abstract—A handful of proposals address the problem of of computers starting up at the same time. In fact, a large bootstrapping a large DHT network from scratch, but they fraction of computers connected to the Internet today routinely all forgo the standard DHT join protocols in favor of their reboot at the same time when they receive a periodic operating own distributed algorithms that build routing tables directly. Motivating their algorithms, the proposals make a perfunctory systems update. In August 2007, one such massive reboot claim that the standard join protocols are not designed to handle triggered a hidden bug in the Skype client software, causing the huge number of concurrent join requests involved in such a world-wide failure of the Skype voice-over-IP network for a bootstrapping scenario. Moreover, the proposals assume a two days [2]. A DHT-based overlay multicast system such as pre-existing unstructured overlay as a starting point for their SplitStream [3] is another example. A large number of users algorithms. We find the assumption somewhat unrealistic. We take a step back and reexamine the performance of the may join the overlay for a live streaming of a popular event. standard DHT join protocols. Starting with nothing other than A few proposals address the problem of bootstrapping a a well-known bootstrap server, when faced with a large number DHT from scratch [4]–[8], but they all forgo the standard DHT of nodes joining nearly simultaneously, can the standard join join protocols in favor of their own distributed algorithms that algorithms form a stable DHT overlay? If so, how quickly? Our build routing tables directly.
    [Show full text]
  • One Ring to Rule Them All: Service Discovery and Binding in Structured Peer-To-Peer Overlay Networks
    One Ring to Rule them All: Service Discovery and Binding in Structured Peer-to-Peer Overlay Networks Miguel Castro Peter Druschel Anne-Marie Kermarrec Antony Rowstron Microsoft Research, Rice University, Microsoft Research, Microsoft Research, 7 J J Thomson Close, 6100 Main Street, 7 J J Thomson Close, 7 J J Thomson Close, Cambridge, MS-132, Houston, Cambridge, Cambridge, CB3 0FB, UK. TX 77005, USA. CB3 0FB, UK. CB3 0FB, UK. [email protected] [email protected] [email protected] [email protected] One Ring to rule them all. One Ring to find them. One Ring to bring In these systems a live node in the overlay to each key them all. And in the darkness bind them. J.R.R. Tolkien and provide primitives to send a message to a key. Mes- Abstract sages are routed to the live node that is currently responsi- ble for the destination key. Keys are chosen from a large space and each node is assigned an identifier (nodeId) cho- Self-organizing, structured peer-to-peer (p2p) overlay sen from the same space. Each node maintains a routing networks like CAN, Chord, Pastry and Tapestry offer a table with nodeIds and IP addresses of other nodes. The novel platform for a variety of scalable and decentralized protocols use these routing tables to assign keys to live distributed applications. These systems provide efficient nodes. For instance, in Pastry, a key is assigned to the live and fault-tolerant routing, object location, and load bal- node with nodeId numerically closest to the key. ancing within a self-organizing overlay network.
    [Show full text]
  • Topology-Aware Routing in Structured Peer-To-Peer Overlay Networks
    Topology-aware routing in structured peer-to-peer overlay networks ¡ ¢ Miguel Castro Peter Druschel Y. Charlie Hu Antony Rowstron Microsoft Research, 7 J J Thomson Close, Cambridge, CB3 0FB, UK. ¡ Rice University, 6100 Main Street, MS-132, Houston, TX 77005, USA. ¢ Purdue University, 1285 EE Building, West Lafayette, IN 47907, USA. Abstract tant distinction lies in the approach they take to consider- ing and exploiting proximity in the underlying Internet. Structured peer-to-peer (p2p) overlay networks like Chord in its original design, for instance, does not con- CAN, Chord, Pastry and Tapestry offer a novel platform sider network proximity at all. As a result, its protocol for a variety of scalable and decentralized distributed ap- for maintaining the overlay network is very light-weight, plications. They provide efficient and fault-tolerant rout- but messages may travel arbitrarily long distances in the ing, object location and load balancing within a self- Internet in each routing hop. organizing overlay network. One important aspect of In a version of CAN, each node measures its network these systems is how they exploit network proximity in delay to a set of landmark nodes, in an effort to deter- the underlying Internet. We present a study of topology- mine its relative position in the Internet and to construct aware routing approaches in p2p overlays, identify prox- an Internet topology-aware overlay. Tapestry and Pastry imity neigbor selection as the most promising technique, construct a topology-aware overlay by choosing nearby and present an improved design in Pastry. Results ob- nodes for inclusion in their routing tables.
    [Show full text]